OpenAI is taking a significant stride towards its stated ambition of creating a unified AI superapp, integrating its flagship ChatGPT with specialized tools like the Codex coding agent, the Atlas browser, and new agentic AI capabilities tailored for the enterprise. The initial manifestation of this grand vision is a substantial update to the Codex desktop application, which is positioned to serve as the foundational platform for this forthcoming integrated AI experience. This evolution signifies OpenAI’s strategic pivot towards consolidating its diverse AI offerings into a more cohesive and powerful ecosystem.
The journey of Codex began in May 2025 as OpenAI’s dedicated software engineering assistant. While its initial adoption was gradual, the platform experienced a dramatic surge in popularity and utility throughout 2026, particularly following the February release of specialized Codex models meticulously trained for developers. This strategic enhancement has propelled Codex to unprecedented user adoption. Thibault Sottiaux, OpenAI’s Head of Codex, revealed in a pre-launch press briefing that the platform now boasts an impressive three million weekly active users, with an accelerating growth rate of one million new users per month.
"We’re actually doing the sneaky thing where we’re building the superapp out in the open—and evolving it out of Codex," Sottiaux remarked, underscoring the transparent, iterative approach OpenAI is taking in developing this integrated AI environment. The Codex desktop app, by its very nature, represents the most logical nexus for this ambitious project. While the recent updates are heavily framed around coding and developer workflows, the underlying functionalities are poised to offer significant advantages to a broader audience, encompassing business professionals and individual users alike.
Background and Evolution of Codex
Codex’s genesis as a coding assistant was rooted in OpenAI’s recognition of the growing demand for AI-powered development tools. Launched in mid-2025, its initial purpose was to streamline the software development lifecycle by assisting with code generation, debugging, and documentation. The platform’s capabilities were gradually enhanced through iterative updates and model refinements.
A pivotal moment in Codex’s trajectory occurred in February 2026 with the introduction of dedicated models fine-tuned for developer tasks. This move signaled OpenAI’s commitment to empowering software engineers, providing them with more sophisticated and context-aware AI assistance. The subsequent user growth, as highlighted by Sottiaux, demonstrates the success of this strategy, transforming Codex from a niche developer tool into a mainstream AI platform with rapidly expanding influence. The current user base of three million weekly users and a monthly growth of one million signifies a powerful network effect, with each new user potentially contributing to the platform’s overall utility and appeal.
Computer Use in the Background: The Dawn of Agentic Operations
A groundbreaking feature introduced in the latest Codex update is its ability to operate desktop applications, currently exclusive to macOS. This capability allows Codex to interact with and control installed software, a functionality that extends far beyond its original coding purview. Whether it’s for testing code generated by its AI agents or automating a diverse range of other tasks, Codex can now perform actions directly on the user’s computer.
Crucially, this feature operates seamlessly in the background, a significant departure from many existing AI agents that often dominate the user interface, potentially hindering productivity. "You stop feeling like there is a limit to what Codex can do," Sottiaux elaborated. "It’s getting very, very creative. It keeps you in the flow. And there’s this understanding of your goals, your context, the ability to work across all of the apps that you have, your browser, and just really staying in lockstep. But also the ability to perform all of this work on a regular basis, in the background, and just really coming in and being useful and helpful on everything that matters to you day-to-day, not just as you’re building and developing software."
This background operation is not merely a convenience; it represents a fundamental shift in how users can leverage AI. Instead of being an active participant, the AI becomes an omnipresent, efficient assistant. OpenAI is even leveraging this capability internally for quality assurance, with Codex agents verifying the implementation of new features within the application itself. "It’s literally Codex building itself," Sottiaux stated, hinting at the self-improving nature of this advanced agentic system. To enable this feature, users are required to download a specific plugin via the Codex settings.
A Browser Integrated Within Codex: Navigating the Web with AI
The integration of OpenAI’s Atlas browser into Codex marks another significant step towards the superapp vision. While the rollout is currently in its nascent stages, the implications are far-reaching, extending well beyond the immediate application previews developers can create. Users can now annotate content directly within the browser interface, a feature particularly beneficial for providing feedback on front-end development work.
Looking ahead, Codex will gain full access to the Atlas browser’s capabilities, enabling it to navigate web pages, execute user flows, and critically analyze its own web-based operations. This deep integration promises to unlock unprecedented potential for developers, offering a streamlined environment for testing and debugging. For business users, the implications are equally profound, with the potential for automated form filling, data extraction between web applications, and more efficient online task management. Sottiaux emphasized that this integration is a preview of future advancements, stating, "It’s not the full power of it yet. It’s kind of a sneak peek of what’s to come there."
Image Generation Within Codex: Enhancing Creative Workflows
Complementing its expanded web and desktop capabilities, Codex now incorporates access to OpenAI’s gpt-image-1.5 image generation model. This feature directly addresses the growing need for visual assets in digital projects. Developers can readily integrate visual elements into websites generated within Codex, or utilize them as placeholders during the development process. Beyond coding applications, the image generation capability is also well-suited for creating mockups and generating assets for games, opening up new creative avenues for users. OpenAI has confirmed that image generation will be provided at no additional cost, further enhancing the value proposition of the updated Codex platform.
Plugins Galore: Expanding Functionality Through Integration
The latest release introduces support for an additional 90 plugins, which are curated bundles of skills, integrations, and MCP servers designed to extend Codex’s capabilities. Notable new additions include integrations with Atlassian Rovo for JIRA management, CircleCI, CodeRabbit, GitLab Issues, Microsoft Suite, Neon by Databricks, Remotion, Render, and Superpowers. Andrew Ambrosio, who leads the development of the Codex app, articulated the strategic intent behind this expansive plugin ecosystem: "the goal is for Codex to fit in naturally with the way that people work, and not make them change how they work."
These plugins are critical for enabling Codex’s non-coding use cases, moving beyond traditional office productivity and email services. The expanded plugin library now includes integrations with platforms like MyRegistry.com, United Rentals, FINN car subscriptions, Readwise, and various research tools from industry leaders such as Pitchbook, Morningstar, and Scite. The true power of these plugins is amplified when combined with Codex’s Automations, a feature that allows prompts to be executed on a schedule, mirroring functionalities like Routines in Claude Code.
Heartbeats and Automations: Intelligent Task Management
The Automations feature has undergone a significant overhaul, drawing inspiration from the OpenClaw framework. The introduction of "heartbeats" allows for the automation of follow-ups within designated threads. "You can set up a thread that’s designed for a daily brief, or a thread that’s designed to triage Slack or something like that," Ambrosio explained. "Or you can set this up as a personal assistant that every five minutes gets a heartbeat run, similar to an Open Claw, and then goes and does a bunch of stuff with plugins. So this is a really powerful axis of personalization for the types of tasks you need to get done, especially those that are pretty persistent."
The persistence of these heartbeats within the same thread is a key differentiator. Competitors often create new worktrees for each automated task, leading to a loss of context from previous iterations. Within OpenAI, these heartbeat automations have proven instrumental in deploying continuously operating agents that monitor Slack channels, triage incoming communications, and track updates across platforms like GitHub and Notion. Sottiaux described running multiple heartbeat automations concurrently, effectively creating AI-powered teammates.
More Proactive Help: Anticipating User Needs
Codex is also becoming more proactive in suggesting follow-up actions, including the scheduling of automations. "Codex now also proactively proposes useful work to continue where you have left off. Using context from projects, connected plugins, and memory, Codex can now suggest how to start your work day or where to pick up on a previous project," the OpenAI team announced. This proactive assistance could manifest as suggestions to follow up on unanswered comments in a shared document, drawing context from integrated applications like Slack and Notion.
Combined with an enhanced memory system that better retains user preferences and contextual information from past interactions, these proactive suggestions represent a significant advancement for day-to-day usability, extending far beyond the traditional developer workflow. This evolution clearly signals OpenAI’s long-term strategy for the Codex application.
While the emphasis on expanding beyond coding is evident, numerous enhancements and quality-of-life improvements are also being rolled out for developers. "We are continuing to invest in places that developers already spend time," Ambrosio stated. For instance, developers can now directly address GitHub review comments within Codex. The platform now supports multiple terminal tabs, an alpha version of SSH-based remote development box connections, and the ability to open various document types (PDFs, spreadsheets, slides) in a sidebar. A new summary pane also provides a consolidated view of plans, sources, and artifacts. "Together, these improvements make it faster to move across all the stages of the software development lifecycle between writing code, checking outputs, reviewing changes, and collaborating with the agent in one workspace," the OpenAI team elaborated.
Codex as a Teammate: Beyond Code Generation
Internally at OpenAI, Codex has transcended its origins as a coding tool, with over 80% of the company utilizing it across a wide spectrum of tasks. These use cases extend beyond engineering, encompassing the drafting of weekly briefs, synthesizing feedback, formulating product requirements, reviewing contracts, and even sending out security training reminders. This widespread adoption underscores Codex’s transformation into a versatile organizational assistant.
Availability and Future Outlook
The updated Codex desktop app is now accessible to all Codex users logged in with their ChatGPT accounts. However, certain advanced personalization features, such as context-aware suggestions and memory enhancements, will be progressively rolled out to Enterprise, Edu, and users in the EU and UK at a later stage. Computer use functionality will also be introduced to EU and UK users soon after the initial launch.
Sottiaux concluded by outlining the future roadmap, indicating that a Codex web app and a mobile application are also planned for future releases. This suggests a comprehensive rollout strategy aimed at making the unified AI superapp accessible across all major platforms, further solidifying OpenAI’s position at the forefront of AI development and integration. The continued evolution of Codex, from a coding assistant to a foundational element of a broader AI superapp, signifies a transformative phase in how individuals and organizations will interact with artificial intelligence in the years to come.